Fig. 3: Examples of the large-weight features in the discriminative model for PSMB5 mutant cells.
From: Machine learning approach for discrimination of genotypes based on bright-field cellular images
a Schematics of the feature extraction procedure from BF images. Tau indicates a threshold of pixel value to identify clumps within a cellular region as described previously15. b Distribution of PSMB5 mutant cells and wild-type cells on the feature “sample mean of number of clumps”. Blue and red histograms represent mutant and wild-type cells, respectively. P value of one-sided U test between mutant and wild-type cells, and the regression coefficient (RC) in the logistic regression model are shown. c Distribution of PSMB5 mutant cells and wild-type cells on the feature “sample mean of average size of clump areas” shown in the same way as in b. d Distribution of PSMB5 mutant cells and wild-type cells on the feature “sample mean of average value of noncircularity” shown in the same way as in b. e Example of clump detection using an image of a wild-type cell. The leftmost panel is an input cellular image, and the right panels are those processed with different thresholds. In the magnified panel, each clump is shown in a different color.
